Add Hugging Face integration to docs

docs/source/index.rst

@@ -222,8 +222,9 @@ Utils
     Definition of helper functions and classes
 
     * :doc:`Utilities <modules/skrl.utils.utilities>`, e.g. setting the random seed
-    * :doc:`Model instantiators <modules/skrl.utils.model_instantiators>`
     * Memory and Tensorboard :doc:`file post-processing <modules/skrl.utils.postprocessing>`
+    * :doc:`Model instantiators <modules/skrl.utils.model_instantiators>`
+    * :doc:`Hugging Face integration <modules/skrl.utils.huggingface>`
     * :doc:`Isaac Gym utils <modules/skrl.utils.isaacgym_utils>`
     * :doc:`Omniverse Isaac Gym utils <modules/skrl.utils.omniverse_isaacgym_utils>`
 
@@ -233,7 +234,8 @@ Utils
     :hidden:
 
     modules/skrl.utils.utilities
-    modules/skrl.utils.model_instantiators
     modules/skrl.utils.postprocessing
+    modules/skrl.utils.model_instantiators
+    modules/skrl.utils.huggingface
     modules/skrl.utils.isaacgym_utils
     modules/skrl.utils.omniverse_isaacgym_utils

docs/source/intro/data.rst

@@ -272,6 +272,30 @@ The following code snippets show how to load the checkpoints through the instant
             # Load the checkpoint
             policy.load("./runs/22-09-29_22-48-49-816281_DDPG/checkpoints/2500_policy.pt")
 
+In addition, it is possible to load, through the library utilities, trained agent checkpoints from the Hugging Face Hub (`huggingface.co/skrl <https://huggingface.co/skrl>`_). See the :doc:`Hugging Face integration <../modules/skrl.utils.huggingface>` for more information.
+
+.. tabs::
+
+    .. tab:: Agent (from Hugging Face Hub)
+
+        .. code-block:: python
+            :emphasize-lines: 2, 13-14
+
+            from skrl.agents.torch.ppo import PPO
+            from skrl.utils.huggingface import download_model_from_huggingface
+
+            # Instantiate the agent
+            agent = PPO(models=models,  # models dict
+                        memory=memory,  # memory instance, or None if not required
+                        cfg=agent_cfg,  # configuration dict (preprocessors, learning rate schedulers, etc.)
+                        observation_space=env.observation_space,
+                        action_space=env.action_space,
+                        device=env.device)
+
+            # Load the checkpoint from Hugging Face Hub
+            path = download_model_from_huggingface("skrl/OmniIsaacGymEnvs-Cartpole-PPO")
+            agent.load(path)
+
 Migrating external checkpoints
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 

docs/source/modules/skrl.utils.huggingface.rst

@@ -0,0 +1,20 @@
+Hugging Face integration
+========================
+
+.. raw:: html
+
+   <hr>
+
+Download model from Hugging Face Hub
+------------------------------------
+
+Several skrl-trained models (agent checkpoints) for different environments/tasks of Gym/Gymnasium, Isaac Gym, Omniverse Isaac Gym, etc. are available in the Hugging Face Hub
+
+These models can be used as comparison benchmarks, for collecting environment transitions in memory (for offline reinforcement learning, e.g.) or for pre-initialization of agents for performing similar tasks, among others
+
+Visit the `skrl organization on the Hugging Face Hub <https://huggingface.co/skrl>`_ to access publicly available models!
+
+API
+"""
+
+.. autofunction:: skrl.utils.huggingface.download_model_from_huggingface

docs/source/modules/skrl.utils.utilities.rst

@@ -2,7 +2,7 @@ Utilities
 =========
 
 .. contents:: Table of Contents
-   :depth: 2
+   :depth: 1
    :local:
    :backlinks: none